ABSTRACT
Artificial intelligence techniques, such as neural network models, fuzzy logic, expert systems, and genetic algorithms, are expected to play important roles in the future sophisticated traffic control systems: they could adjust traffic signal timings automatically in response to traffic situations. This chapter shows that the development of input-output relationship between the signal timings and the objective function using a multilayer neural network model decreases computational burdens because simulations are not necessary in each optimization step. It focuses on how people can apply neural network models, such as a multilayer neural network (NN) model and a Kohonen Feature Map (KFM) model, to the subjects concerning traffic flow problems. The chapter present conceptual flow diagrams of how traffic signal timings are trained and optimized using the models. It also presents a procedure of how to describe the two or three dimensional relationship using neural models.
